Childhood Trauma, Personality Traits, Self-Efficacy, and Relationship Dependency in the Context of Online Dating Application Use

Abstract

Although online dating applications have assumed an increasingly central role in the formation of romantic and interpersonal relationships, research explaining the use of these applications by jointly considering adverse early-life experiences and psychological characteristics in adulthood remains limited. This study aimed to examine whether adults who use and do not use online dating applications differ in terms of childhood trauma, Big Five personality traits, general self-efficacy, and relationship dependency, and to determine the relative and incremental associations of these variables with application use. The study employed a cross-sectional and correlational design and included 431 adults aged 18–57 years (M_age = 28.24, SD = 7.95); 199 participants (46.2%) reported using online dating applications, whereas 232 (53.8%) reported not using them. The groups were compared using chi-square tests, independent-samples t-tests, MANOVA, and MANCOVA; variables independently and incrementally associated with application use were examined using hierarchical binary logistic regression. Participants who used online dating applications had markedly higher total childhood trauma scores than non-users (d = 1.37), and significant group differences were identified across all trauma subdimensions. The largest unadjusted group difference emerged for childhood sexual abuse (η²p = 0.343). Users had lower levels of general self-efficacy (d = 0.52), and significant group differences were found across all Big Five personality traits, whereas no significant difference was observed in relationship dependency. After controlling for age, gender, economic status, and educational level, the overall multivariate group difference remained significant (Pillai’s Trace = 0.212, F(12, 409) = 9.16, p < 0.001, η²p = 0.212). In the hierarchical logistic regression, childhood trauma made a significant incremental contribution to the model beyond demographic variables. In the final model, gender and childhood sexual abuse remained independently associated with application use; each one-unit increase in the sexual abuse score was associated with increased odds of application use (OR = 1.217, 95% CI [1.114, 1.330], p < 0.001). Self-efficacy showed a more limited independent association (OR = 0.966, p = 0.048), whereas personality traits and relationship dependency were not independently significant. Gender-stratified analyses substantially qualified the pooled trauma findings: the total-trauma difference was nonsignificant among women (d = 0.25, p = 0.134) but large among men (d = 1.78, p < 0.001), while the sexual-abuse difference was present in both genders but substantially larger among men. The findings therefore indicate sample-specific associations with meaningful gender heterogeneity and should not be interpreted as causal, temporal, or representative of particular patterns or motivations of application use.

1. Introduction

Technological transformation has substantially changed not only interpersonal communication practices but also the ways in which romantic and sexual relationships are initiated, maintained, and terminated. As one of the most visible manifestations of this transformation, online dating applications reduce geographical and temporal constraints, allowing users to access a broad pool of potential partners, make selections based on profiles, and communicate prior to face-to-face encounters (Aydoğan, 2020; Heino et al., 2010; LeFebvre, 2018). Recent systematic reviews also indicate that mobile dating applications have become an important component of contemporary romantic and sexual relationship formation, while patterns of use and their psychological correlates remain highly heterogeneous (Castro & Barrada, 2020; Konings et al., 2022).

Online dating application use cannot be explained by a single motivation. In addition to seeking romantic partners or sexual encounters, interpersonal and intrinsic motivations such as socializing, entertainment, curiosity, ease of communication, self-worth validation, sensation seeking, and social acceptance are also associated with application use (Ranzini & Lutz, 2017; Sumter et al., 2017; Van De Wiele & Tong, 2014). Findings from systematic reviews suggest that motivations for use may be differentially associated with gender, sexual orientation, sociosexuality, sensation seeking, personality, and interpersonal characteristics (Konings et al., 2022). Accordingly, online dating application use should not be regarded as a homogeneous behavior but rather as a multidimensional behavioral pattern that interacts with individual characteristics and the social context.

Findings concerning the demographic characteristics of individuals who use these applications are also not entirely consistent. Castro and Barrada (2020), in their systematic review, reported that demographic characteristics such as age, gender, sexual orientation, and relationship status may be associated with online dating application use, whereas Holtzhausen et al. (2020) identified associations with younger age and sexual minority status but found no significant difference by gender. Shapiro et al. (2017), in turn, demonstrated that educational level, sexual needs, and relationship status were associated with Tinder use. This heterogeneity highlights the importance of considering potential confounding variables such as age, gender, education, and socioeconomic characteristics when examining the psychological correlates of application use.

Similarly, associations between online dating application use and psychological health and well-being have received increasing research attention. Previous studies have suggested that use may be associated with depression, anxiety, stress, body image, and self-related variables (Blackhart et al., 2014; Holtzhausen et al., 2020; Lo & Chui, 2024). More recent systematic review findings indicate that dating application use may be associated particularly with body image, psychological distress, and indicators of well-being; however, because existing studies are predominantly cross-sectional, the direction of these associations and the underlying mechanisms remain unclear (Bowman et al., 2025). A 2026 systematic review and meta-analysis likewise reported small-to-moderate differences between users and non-users in emotional distress, body image concerns, behavioral dysregulation, and interpersonal sensitivity, although the high heterogeneity across studies points to the importance of individual predispositions and social context.

An important methodological problem in the literature is that online dating application use is sometimes directly equated with “problematic use” or adverse psychological outcomes. However, using an application and problematic use are distinct constructs. Recent systematic reviews indicate that a substantial proportion of existing studies do not adequately distinguish conceptually between motivations for use, problematic use, and psychological outcomes, and that research is largely based on single-time-point cross-sectional data (Thomas et al., 2025). Therefore, rather than regarding application use itself as an indicator of psychopathology or dysfunctional behavior, examining developmental and psychological characteristics that may be associated with use in terms of their independent contributions offers a more cautious approach.

Within this framework, adverse early-life experiences are particularly noteworthy. Although childhood trauma is associated with self-perception, emotion regulation, interpersonal trust, intimacy, and romantic relationship patterns in adulthood, the online dating application literature has predominantly focused on user motivations, personality, sexual behaviors, psychological well-being, and application-related risks. Studies that examine childhood trauma, personality traits, self-efficacy, and relationship dependency within the same analytical framework and evaluate their relative contributions beyond demographic characteristics remain limited. This gap necessitates a more comprehensive developmental and psychological examination of individual differences in online dating application use.

1.1. Childhood Trauma and Online Dating Application Use

Childhood trauma encompasses adverse early-life experiences such as emotional, physical, and sexual abuse, as well as emotional and physical neglect. The effects of these experiences may not be confined to childhood and may be associated with self-evaluations, interpersonal trust, intimacy, emotion regulation, and romantic relationship patterns in adulthood. Recent findings showing that childhood maltreatment is associated with risky behaviors and victimization in later romantic relationships also support this developmental continuity (McIntyre et al., 2024).

Exposure to maltreatment during childhood has been associated with negative self-evaluations, shame, self-criticism, and various psychological difficulties (Miguel et al., 2025). The association of such developmental experiences with interpersonal intimacy and trust processes suggests that online romantic interactions may serve different psychological functions for some individuals. The fact that online environments do not initially require physical proximity, provide relatively greater control over the timing of communication, and allow interactions with potential partners to be initiated gradually may be important for individuals who have experienced early interpersonal adversity. However, this potential mechanism cannot be directly inferred from the present cross-sectional data and requires empirical testing.

The emerging literature on the association between childhood maltreatment and online dating experiences has focused particularly on revictimization. Fereidooni et al. (2022) reported that, among young adult women using mobile dating applications, childhood maltreatment was positively associated with the severity of both online and face-to-face sexual revictimization and that motivations involving regulation of negative affect and enhancement of self-esteem played a role in some of these associations. A systematic scoping review examining sexual harassment in dating-app environments also identified early childhood maltreatment as one of the psychosocial factors associated with app-facilitated sexual harassment experiences (Gewirtz-Meydan et al., 2024).

Recent findings regarding the role of childhood sexual abuse in the context of online dating are particularly noteworthy. In a study of 534 dating application users, Pampanel et al. (2024) reported that individuals with a history of childhood sexual abuse experienced greater exposure to technology-facilitated sexual violence through applications; furthermore, these individuals showed lower love motivations and higher motivations related to self-worth validation, ease of communication, and sensation seeking. These findings indicate that understanding the relationship between early sexual trauma and digital romantic interactions requires attention not only to the variable of “use” but also to motivations for use and the nature of online experiences.

Nevertheless, most existing studies examine victimization, risky experiences, or motivations for use among individuals who are already using online dating applications. Therefore, directly comparing different dimensions of childhood trauma between adult users and non-users and determining whether these differences persist after demographic and other psychological variables are taken into account addresses an important gap in the literature.

1.2. Personality Traits and Online Dating Application Use

Personality is one of the constructs most frequently examined to explain individual differences in online dating application use. Within the framework of the Five-Factor Model of personality, extraversion, agreeableness, conscientiousness, neuroticism, and openness to experience provide a theoretically meaningful framework for understanding social interaction preferences, novelty seeking, interpersonal approach, and digital communication behaviors.

However, empirical findings concerning the relationship between personality and online dating application use are inconsistent. Some studies have reported differences in certain personality traits between users and non-users (Fowler & Both, 2020; Timmermans & De Caluwé, 2017). In contrast, other studies have found that personality traits do not strongly or consistently distinguish application use (Castro et al., 2020; Freyth de Polo León & Batinic, 2021). Systematic review findings also suggest that personality may be differentially associated with specific motivations for use rather than with general application use; for example, extraversion appears to be associated with some relational motives for use, whereas certain characteristics that impair social functioning have been linked to intrinsic motivations (Konings et al., 2022).

This heterogeneity indicates that treating the association between personality traits and application use as an independent and invariant explanation may be problematic. It is possible that observed user–non-user differences in personality are partly attributable to demographic variables such as age, gender, and socioeconomic characteristics or to early developmental experiences. Therefore, evaluating the relationship between personality and online dating application use within the same multivariable model as childhood trauma and other psychological characteristics may allow the independent contribution of personality to be determined more accurately.

1.3. Self-Efficacy and Online Dating Application Use

Self-efficacy refers to an individual’s belief in their capacity to perform the behaviors required in specific situations and occupies a central position within social cognitive theory (Bandura, 1989). General self-efficacy reflects a broader belief in one’s capacity to cope with different and challenging situations (Yıldırım & İlhan, 2010). Thus, self-efficacy concerns not so much the skills an individual objectively possesses as their appraisal of whether they can use those skills effectively.

Online dating environments differ from face-to-face romantic interactions in several respects. The ability of users to manage their profiles, decide when to initiate or terminate communication, and select among potential partners may provide relatively greater control over the interaction process. Therefore, individuals with lower levels of general self-efficacy may perceive online environments as more manageable than face-to-face romantic interactions. However, this possible explanation should be considered with the recognition that general self-efficacy and social or romantic self-efficacy are not the same construct.

A substantial proportion of the existing online dating literature has focused on self-esteem, self-presentation, body image, and motivations for use, whereas the relationship between general self-efficacy and application use has received comparatively less attention. Recent systematic reviews likewise indicate that variables related to self-attitudes have been examined, but that conceptual and measurement heterogeneity remains high. Therefore, examining whether general self-efficacy differs between adult users and non-users and whether it demonstrates an independent association beyond demographic, developmental, and personality characteristics may provide a complementary contribution to the literature.

1.4. Relationship Dependency and Online Dating Application Use

Relationship dependency is characterized by patterns such as focusing on the needs of others rather than one’s own needs, assuming excessive responsibility for others, difficulty setting boundaries, difficulty expressing emotions, and regulating self-worth substantially through interpersonal relationships (Fischer et al., 1991). Because of these characteristics, relationship dependency represents a potentially meaningful variable in assessing individual differences in the establishment and maintenance of close relationships.

The ability of online dating applications to facilitate access to new romantic or sexual partners, accelerate the search for relationships, and provide users with a broad pool of potential partners may suggest a possible association between relationship dependency and application use. However, relationship dependency is primarily defined through behavioral and emotional patterns within established interpersonal relationships, whereas online dating applications may be used for many different motivations. Therefore, merely using a dating application cannot be assumed to indicate high relationship dependency.

Indeed, the recent literature indicates that online dating behaviors may be undertaken for diverse purposes, ranging from seeking romantic relationships to entertainment, and from sexuality to self-worth validation. Moreover, the problematic online dating literature has been reported to sometimes conceptually conflate use behavior with the adverse consequences of use (Thomas et al., 2025). Therefore, whether relationship dependency actually demonstrates an independent association with application use needs to be tested while controlling for other psychological and demographic characteristics.

1.5. The Present Study

The existing literature indicates that online dating application use is associated with demographic characteristics, personality, psychological health, self-evaluations, motivations for use, and interpersonal outcomes. The present study does not propose a single causal mechanism linking the selected constructs. Rather, it uses a developmental-individual-differences framework in which the domains are conceptually differentiated by level: childhood trauma represents adverse early interpersonal and developmental experience; Five-Factor traits represent broad dispositional tendencies; general self-efficacy represents a generalized appraisal of coping capability; and relationship dependency represents an interpersonal-relational tendency. Examining these domains together therefore allows the analysis to ask whether information from developmentally earlier experiences, broad dispositional characteristics, generalized self-appraisal, and relationship-oriented tendencies contributes differently to binary application-use status once demographic context is taken into account. The hierarchical order was chosen to reflect this comparative logic rather than a causal sequence: demographic characteristics were entered first as contextual controls, childhood-trauma dimensions were then added as developmental-history variables, followed by personality traits, general self-efficacy, and relationship dependency as progressively more proximal adult individual-difference and relational domains. Thus, the model tests relative and incremental statistical associations across complementary domains; it does not test mediation, a developmental pathway, or an explanatory mechanism linking childhood trauma to application use.

Three empirical gaps motivate this approach. First, a substantial proportion of studies focus only on current application users, whereas direct user–non-user comparisons, particularly regarding early developmental experiences, remain limited. Second, research connecting childhood trauma with online dating has largely focused on revictimization and technology-facilitated sexual violence (Fereidooni et al., 2022; Gewirtz-Meydan et al., 2024; Pampanel et al., 2024). Third, relatively few studies have evaluated developmental history, broad personality, generalized self-efficacy, interpersonal dependency, and demographic context within the same model. The Turkish context also broadens the cultural scope of the literature in which gender norms, relationship expectations, and sociocultural context may shape online romantic behavior. Importantly, because application use was measured only as user versus non-user, the present study addresses use status rather than frequency, duration, intensity, platform, motivation, or problematic use.

Accordingly, the aim of the present study was to examine whether adults classified as online dating application users and non-users differed in childhood trauma, Five-Factor personality traits, general self-efficacy, and relationship dependency, and to evaluate the relative and incremental statistical associations of these domains with binary application-use status beyond measured demographic characteristics. The study also examined whether different childhood trauma dimensions carried distinct statistical information within the multivariable model. No mediation, developmental pathway, or causal mechanism was tested.

2. Materials and Methods

2.1. Research Design

The study was conducted using a correlational survey model, one of the quantitative research methods (Bekman, 2022). The study was cross-sectional and observational, with online dating application use status treated as the primary outcome variable. The relationships between childhood trauma, Five-Factor personality traits, general self-efficacy, relationship dependency, and online dating application use were examined while taking sociodemographic characteristics into account. Because of the cross-sectional design, the terms “prediction” and “predictor” used in the analyses refer solely to statistical modeling and do not imply causal or temporal relationships. The study was reported in accordance with the STROBE guidelines for cross-sectional observational studies.

2.2. Participants and Sampling

Convenience and snowball sampling, both non-probability sampling methods, were used to recruit the study sample. The sole eligibility criterion was age: participants were required to be 18–65 years old. Online dating application use itself was not an eligibility criterion because the study compared users and non-users.

A total of 431 adults aged 18–57 years participated in the study. The mean age of the participants was 28.24 years (SD = 7.95). Of the participants, 229 (53.1%) were women and 202 (46.9%) were men. A total of 199 participants (46.2%) reported using online dating applications, whereas 232 (53.8%) reported not using them. Because all questions in the research form were set as mandatory, there were no missing data, and all 431 participants were included in the analyses (Table 1).

Table 1. Demographic Characteristics of Participants by Online Dating Application Use Status.

Demographic Variable	Users, n (%)	Non-Users, n (%)	χ2	p	Effect Size
Gender			138.28	<0.001	φ = 0.57
Women	45 (19.7)	184 (80.3)
Men	154 (76.2)	48 (23.8)
Age Group			27.04	<0.001	V = 0.25
Emerging adulthood	166 (53.2)	146 (46.8)
Established adulthood	31 (32.6)	64 (67.4)
Middle adulthood	2 (8.3)	22 (91.7)
Educational Level			7.42	0.060	V = 0.13
Primary education or below	12 (34.3)	23 (65.7)
High school	64 (48.5)	68 (51.5)
University	92 (43.2)	121 (56.8)
Postgraduate	31 (60.8)	20 (39.2)
Economic Status			21.26	<0.001	V = 0.22
Low	43 (60.6)	28 (39.4)
Middle	132 (40.2)	196 (59.8)
High	24 (75.0)	8 (25.0)
Employment Status			20.37	<0.001	V = 0.22
Employed	126 (55.8)	100 (44.2)
Unemployed	34 (37.8)	56 (62.2)
Retired	1 (10.0)	9 (90.0)
Student	38 (36.2)	67 (63.8)
Relationship Status			25.69	<0.001	V = 0.24
In a relationship	58 (36.7)	100 (63.3)
Not in a relationship	85 (64.4)	47 (35.6)
Married	56 (39.7)	85 (60.3)

Note. N = 431. Percentages are row percentages. φ = phi coefficient; V = Cramer’s V.

2.3. Data Collection Instruments

2.3.1. Personal Information Form

The Personal Information Form developed by the researcher was used to determine participants’ gender, age, educational level, perceived economic status, employment status, and relationship status. The Personal Information Form also assessed participants’ self-reported online dating application use at the time of the survey. Participants were classified as users if they reported using online dating applications and as non-users if they reported not using them. Application-use status was coded as a binary variable (0 = non-user; 1 = user) for the statistical analyses. Participants who reported application use were additionally asked about their duration of use, daily time spent on applications, and motivations for use. These supplementary characteristics were not used to define user status or included as predictors in the present analyses. Accordingly, the classification reflects self-reported use status at the time of assessment rather than lifetime history of application use or use within a predefined retrospective period.

2.3.2. Spann–Fischer Codependency Scale

The Spann–Fischer Codependency Scale was developed by Fischer et al. (1991) and adapted into Turkish by Tanhan and Mukba (2014). The scale is a 16-item, single-factor, 6-point Likert-type measure. It assesses characteristics associated with relationship dependency, such as excessive focus on the needs of others, difficulty establishing interpersonal boundaries, assuming excessive responsibility for others, and a tendency to regulate self-worth through relationships. Higher scores indicate higher levels of relationship dependency. In the present study, the Cronbach’s alpha internal consistency coefficient was 0.67. Given this value, findings concerning relationship dependency were interpreted cautiously in terms of measurement reliability.

2.3.3. Childhood Trauma Questionnaire

The Childhood Trauma Questionnaire (CTQ) was developed by Bernstein et al. (1994) to retrospectively assess individuals’ traumatic childhood experiences. The Turkish adaptation was conducted by Şar et al. (2012), and the instrument was subsequently revised by Şar et al. (2021) to create the 33-item CTQ-33.

The CTQ-33 includes the dimensions of emotional abuse, physical abuse, sexual abuse, emotional neglect, physical neglect, and overprotection/overcontrol. In accordance with the aims of the present study, the dimensions of emotional abuse, physical abuse, sexual abuse, emotional neglect, and physical neglect were used; the overprotection/overcontrol dimension was not included in the analyses. The total trauma score was evaluated based on these five subdimensions.

In the present study, Cronbach’s alpha was 0.93 for the total childhood trauma score, 0.80 for emotional abuse, 0.85 for physical abuse, 0.86 for sexual abuse, 0.75 for emotional neglect, and 0.81 for physical neglect.

2.3.4. Short Form of the International Personality Item Pool

The Short Form of the International Personality Item Pool is based on Goldberg’s (1999) item pool developed within the Five-Factor Model of personality. Its Turkish validity and reliability study was conducted by Güneri Yöyen (2016). The scale assesses extraversion, agreeableness, conscientiousness, neuroticism, and openness to experience.

In the present study, the Cronbach’s alpha coefficient for the overall scale was 0.80. Cronbach’s alpha coefficients for the subdimensions were 0.70 for extraversion, 0.70 for agreeableness, 0.73 for conscientiousness, 0.77 for neuroticism, and 0.63 for openness to experience. The relatively low internal consistency coefficient for openness to experience, in particular, was taken into consideration when interpreting findings related to this subdimension.

2.3.5. General Self-Efficacy Scale

The General Self-Efficacy Scale was developed by Sherer et al. (1982) and adapted into Turkish by Yıldırım and İlhan (2010). The Turkish form consists of 17 items and assesses individuals’ general beliefs regarding their capacity to cope with various difficulties. Higher scores indicate higher levels of general self-efficacy. In the present study, the total score rather than the subscale scores was used, and Cronbach’s alpha was calculated as 0.83 (Table 2).

Table 2. Internal Consistency Coefficients of the Measures Used in the Study.

Measure/Subscale	Number of Items	Turkish Adaptation α	Present Study α
Spann–Fischer Codependency Scale (SFCDS)
Total scale	16	0.65	0.67
Childhood Trauma Questionnaire (CTQ)
Total scale	25	0.87	0.93
Emotional abuse	5	0.88	0.80
Physical abuse	5	0.81	0.85
Physical neglect	5	0.77	0.81
Emotional neglect	5	0.89	0.75
Sexual abuse	5	0.90	0.86
Short Form of the International Personality Item Pool (IPIP-SF)
Total scale	40	0.81	0.80
Extraversion	8	0.75	0.70
Conscientiousness	9	0.71	0.73
Agreeableness	8	0.64	0.70
Openness to experience	7	0.65	0.63
Neuroticism	8	0.76	0.77
General Self-Efficacy Scale
Total scale	17	0.80	0.83

Note. α = Cronbach’s alpha internal consistency coefficient. No items were removed from the scales.

2.4. Procedure and Ethics

Approval to conduct the study was obtained from the Maltepe University Ethics Committee on 2 October 2025 (No. 2025/19; Decision No. 2025/19-03). Before participation, all participants were informed about the purpose of the study, that participation was voluntary, that they could withdraw from the study at any time, that the data would be evaluated only for scientific purposes and in aggregate form, and that no identifying information would be collected. Electronic informed consent was obtained from all participants.

Data were collected online via Google Forms between 18 November 2025 and 3 May 2026. The research link was shared through Instagram, WhatsApp, LinkedIn, and Telegram, and participants were asked to forward the research link to individuals who met the inclusion criteria. Because all questions in the data collection form were set as mandatory, there were no missing data, and all 431 participants were included in the analyses.

2.5. Statistical Analysis

Statistical analyses were performed using IBM SPSS Statistics 26.0. Before the analyses, data accuracy, missing data, distributional characteristics, and the internal consistency coefficients of the scales were examined. Because there were no missing data, no data imputation or replacement procedure was applied.

Associations between online dating application use and categorical sociodemographic variables were evaluated using Pearson’s chi-square test of independence. Effect sizes were reported using Phi (φ) for variables with two categories and Cramer’s V for variables with more than two categories. Independent-samples t-tests were used to compare relationship dependency, total childhood trauma, and general self-efficacy scores between participants who used and did not use online dating applications. Homogeneity of variance was evaluated using Levene’s test; when the homogeneity assumption was not met, Welch-adjusted results that did not assume equal variances were used. Effect sizes for between-group differences were reported using Cohen’s d.

Separate multivariate analyses of variance (MANOVAs) were conducted to determine whether the five subdimensions of childhood trauma and the Five-Factor personality traits jointly differed according to online dating application use status. Multivariate effects were reported using Pillai’s Trace; following significant multivariate effects, univariate tests were examined and effect sizes were presented as partial eta squared (η²p).

A multivariate general linear model was applied to determine whether the psychological differences observed between online dating application users and non-users were independent of sociodemographic characteristics. Online dating application use status was entered into the model as the primary fixed factor; age was specified as a continuous covariate, whereas gender, economic status, and educational level were specified as categorical fixed factors. The adjusted main effect of application use status on the psychological dependent variables was evaluated, with the overall multivariate effect reported using Pillai’s Trace and univariate effects reported using partial eta squared (η²p).

Hierarchical binary logistic regression analysis was conducted to evaluate variables independently and incrementally associated with online dating application use status. The dependent variable was coded as 0 = non-user and 1 = user. Demographic variables (age, gender, economic status, educational level, and employment status) were entered in the first block; childhood trauma subdimensions in the second block; Five-Factor personality traits in the third block; general self-efficacy in the fourth block; and relationship dependency in the fifth block. The variable blocks were ordered to examine the incremental statistical contribution of conceptually distinct domains. Demographic characteristics were entered first to establish a baseline model accounting for measured sociodemographic differences between users and non-users. Childhood trauma was subsequently introduced as a retrospectively assessed developmental domain, followed by broad personality traits, general self-efficacy, and relationship dependency as progressively more specific psychological constructs. This ordering was intended to evaluate whether each additional domain provided information beyond the variables already included in the model. It was not intended to represent a causal sequence, establish mediation, or imply that the order of statistical entry reflects the temporal development of online dating application use. The primary outcome was deliberately limited to binary application-use status. Accordingly, the present analyses do not characterize frequency, duration, intensity, platform, motivation, or specific patterns of use, and the findings should not be interpreted as explaining these dimensions of online dating behavior.

The contribution of each model step was evaluated using the likelihood-ratio chi-square change (Δχ²). Model explanatory power was assessed using Cox–Snell R² and Nagelkerke R², and overall model fit was examined using the Hosmer–Lemeshow test. Independent associations in the final model were reported using unstandardized regression coefficients (B), standard errors, Wald χ² statistics, odds ratios (ORs), and 95% confidence intervals. Model classification performance was evaluated using overall accuracy, sensitivity, and specificity. Model discrimination was additionally assessed using the area under the receiver operating characteristic curve (ROC-AUC). As a supplementary internal validation analysis, stratified five-fold cross-validation was conducted using L2-regularized logistic regression (C = 1). Numerical predictors were standardized separately within each training fold to prevent information leakage. Mean ROC-AUC and classification accuracy were calculated across the five held-out folds. This supplementary procedure was distinguished from the original unpenalized hierarchical logistic regression. Multicollinearity in the final predictor set was evaluated using variance inflation factors (VIFs) and tolerance values. Because gender was strongly associated with application-use status in the sample, supplementary sensitivity analyses were conducted to evaluate the robustness of the principal trauma findings. Welch-adjusted user–non-user comparisons of total childhood trauma and the five trauma subdimensions were repeated separately for women and men. In addition, separate general linear models were used to test application-use-status × gender interactions for the central total-trauma and sexual-abuse findings while controlling for age, economic status, and educational level. To further examine the unusually large childhood trauma differences among male participants, supplementary Mann–Whitney U tests and Welch tests based on 20% trimmed means were conducted for total childhood trauma and childhood sexual abuse scores. These analyses were used to assess whether the observed group differences were robust to departures from normality and the influence of extreme observations. All statistical tests were two-tailed, and statistical significance was set at p < 0.05.

3. Results

3.1. Demographic Characteristics

Statistically significant associations were found between online dating application use and gender, age group, economic status, employment status, and relationship status. The association between educational level and application use was not statistically significant. Examination of the effect sizes showed that the strongest association was observed for gender (φ = 0.57), whereas the effect sizes of the other significant associations ranged from small to moderate.

3.2. Relationship Dependency

The mean relationship dependency score of participants who used online dating applications was 56.32 (SD = 9.06), whereas the mean score of non-users was 55.53 (SD = 12.20). The difference between the groups was not statistically significant, t(420.66) = 0.77, p = 0.442, Cohen’s d = 0.07. The negligible effect size indicates that users and non-users were largely similar in terms of relationship dependency.

3.3. Childhood Trauma

The mean total childhood trauma score of participants who used online dating applications (M = 62.01, SD = 18.46) was significantly higher than that of participants who did not use them (M = 39.88, SD = 13.82). Because the assumption of homogeneity of variance was not met, Welch-adjusted results were used, and the group difference was statistically significant, t(362.31) = 13.90, p < 0.001, Cohen’s d = 1.37. The effect size indicated a large difference between the two groups.

When the five trauma subdimensions were considered jointly, the multivariate group effect was significant, Pillai’s Trace = 0.393, F(5, 425) = 55.13, p < 0.001, η²p = 0.393. Univariate analyses showed that participants who used online dating applications had significantly higher scores than non-users across all trauma subdimensions. The largest effect was observed for sexual abuse, followed by physical abuse, emotional abuse, physical neglect, and emotional neglect (Table 3).

Table 3. Comparison of Childhood Trauma Subdimensions by Online Dating Application Use Status.

Subdimension	Users, M (SD)	Non-Users, M (SD)	F(1, 429)	p	η2p
Emotional abuse	12.61 (4.45)	8.13 (3.97)	121.59	<0.001	0.221
Physical abuse	11.36 (5.08)	6.19 (2.89)	174.27	<0.001	0.289
Physical neglect	12.17 (4.92)	7.91 (3.59)	107.25	<0.001	0.200
Emotional neglect	14.03 (3.91)	11.67 (4.89)	29.73	<0.001	0.065
Sexual abuse	11.85 (5.15)	5.97 (2.84)	223.84	<0.001	0.343

Note. Multivariate test: Pillai’s Trace = 0.393, F(5, 425) = 55.13, p < 0.001, η²p = 0.393. η²p = partial eta squared.

3.4. Personality Traits

When the Five-Factor personality traits were considered jointly, a significant multivariate difference was found between participants who used and did not use online dating applications, Pillai’s Trace = 0.293, F(5, 425) = 35.30, p < 0.001, η²p = 0.293.

Univariate analyses revealed significant group differences across all five personality dimensions: extraversion, agreeableness, conscientiousness, neuroticism, and openness to experience. Non-users had higher mean scores than users across all personality dimensions. The largest effect sizes were observed for agreeableness and conscientiousness (Table 4).

Table 4. Comparison of Personality Traits by Online Dating Application Use Status.

Personality Trait	Users, M (SD)	Non-Users, M (SD)	F(1, 429)	p	η2p
Extraversion	12.72 (3.36)	14.12 (3.55)	17.43	<0.001	0.039
Agreeableness	17.15 (3.89)	20.77 (3.08)	116.05	<0.001	0.213
Conscientiousness	16.77 (3.63)	20.14 (3.25)	103.19	<0.001	0.194
Neuroticism	25.53 (6.05)	28.24 (7.02)	18.09	<0.001	0.040
Openness to experience	19.76 (4.16)	22.23 (4.05)	38.81	<0.001	0.083

Note. Multivariate test: Pillai’s Trace = 0.293, F(5, 425) = 35.30, p < 0.001, η²p = 0.293.

3.5. General Self-Efficacy

Participants who used online dating applications had a significantly lower mean general self-efficacy score (M = 54.61, SD = 10.10) than non-users (M = 60.25, SD = 11.54). Because the assumption of homogeneity of variance was not met, Welch-adjusted results were used, t(428.81) = −5.40, p < 0.001, Cohen’s d = 0.52. The effect size was moderate.

3.6. Group Differences After Controlling for Demographic Variables

When age was entered into the model as a continuous covariate and gender, economic status, and educational level as categorical control factors, the overall multivariate difference between online dating application users and non-users remained significant, Pillai’s Trace = 0.212, F(12, 409) = 9.16, p < 0.001, η²p = 0.212.

In the univariate analyses, significant group differences remained for the total childhood trauma score and all trauma subdimensions, all Five-Factor personality traits, and general self-efficacy. In contrast, no significant difference was found between the groups in relationship dependency, F(1, 420) = 0.46, p = 0.499, η²p = 0.001. After controlling for demographic variables, the largest effect was observed for sexual abuse (η²p = 0.155), followed by the total childhood trauma score (η²p = 0.114), physical abuse (η²p = 0.091), and emotional abuse (η²p = 0.080) (Table 5).

Table 5. Multivariate General Linear Model Results for Psychological Variables by Online Dating Application Use Status.

Dependent Variable	F(1, 420)	p	η2p
Relationship dependency	0.46	0.499	0.001
Emotional abuse	36.72	<0.001	0.080
Physical abuse	42.23	<0.001	0.091
Physical neglect	17.79	<0.001	0.041
Emotional neglect	5.71	0.017	0.013
Sexual abuse	77.23	<0.001	0.155
Total childhood trauma score	54.07	<0.001	0.114
Extraversion	3.93	0.048	0.009
Agreeableness	25.99	<0.001	0.058
Conscientiousness	20.00	<0.001	0.045
Neuroticism	4.21	0.041	0.010
Openness to experience	14.55	<0.001	0.033
General self-efficacy	11.30	<0.001	0.026

Note. Age was entered as a continuous covariate; gender, economic status, and educational level were entered as categorical control factors. The overall multivariate effect represents the adjusted main effect of online dating application use status.

3.7. Hierarchical Binary Logistic Regression

A five-step hierarchical binary logistic regression analysis was conducted to evaluate variables independently and incrementally associated with online dating application use.

The first model, consisting of demographic variables, was statistically significant, model χ²(10) = 179.784, p < 0.001, with Nagelkerke R² = 0.456. In the second step, adding the childhood trauma subdimensions significantly improved model fit, Δχ²(5) = 76.242, p < 0.001, and Nagelkerke R² increased to 0.598. Thus, childhood trauma made a substantial incremental contribution to the model beyond demographic variables.

In the third step, adding the Five-Factor personality traits produced a smaller but significant improvement in the model, Δχ²(5) = 12.515, p = 0.028, with Nagelkerke R² = 0.619. In the fourth step, the addition of general self-efficacy also provided a limited but significant incremental contribution, Δχ²(1) = 4.198, p = 0.040, with Nagelkerke R² = 0.626. In the final step, adding relationship dependency did not improve model fit, Δχ²(1) = 0.014, p = 0.907, and Nagelkerke R² remained unchanged at 0.626 (Table 6).

Table 6. Fit and Explanatory Indices of the Hierarchical Binary Logistic Regression Models.

Model	Block Added	Δχ2	df	p	Model χ2	df	p	−2LL	Cox–Snell R2	Nagelkerke R2	H–L χ2	df	p
1	Demographic variables	179.784	10	<0.001	179.784	10	<0.001	415.180	0.341	0.456	3.818	8	0.873
2	Childhood trauma	76.242	5	<0.001	256.026	15	<0.001	338.938	0.448	0.598	9.922	8	0.271
3	Personality traits	12.515	5	0.028	268.541	20	<0.001	326.423	0.464	0.619	7.789	8	0.454
4	General self-efficacy	4.198	1	0.040	272.740	21	<0.001	322.224	0.469	0.626	7.305	8	0.504
5	Relationship dependency	0.014	1	0.907	272.753	22	<0.001	322.210	0.469	0.626	7.273	8	0.507

Note. Δχ² = likelihood-ratio chi-square change from the preceding model; −2LL = −2 log likelihood; H–L = Hosmer–Lemeshow test.

The nonsignificant Hosmer–Lemeshow tests across all models indicated acceptable fit to the observed data. The final model achieved an overall classification accuracy of 82.8%, with a sensitivity of 75.4% and a specificity of 89.2%. Supplementary ROC analysis indicated good discrimination for the final hierarchical logistic regression model (AUC = 0.902). In stratified five-fold cross-validation using L2-regularized logistic regression, the mean ROC-AUC was 0.873 (SD = 0.037), and the mean classification accuracy was 79.12%. These results provide supplementary evidence of internal predictive performance, although the cross-validation procedure used a regularized estimator rather than the original unpenalized model (Table 7).

Table 7. Final Model Results of the Hierarchical Binary Logistic Regression.

Variable	B	SE	Wald χ2	df	p	OR	95% CI
Women (ref. men)	−1.601	0.343	21.788	1	<0.001	0.202	[0.103, 0.395]
Age	−0.056	0.029	3.752	1	0.053	0.945	[0.893, 1.001]
Economic status (overall test)	—	—	1.637	2	0.441	—	—
Low (ref. high)	−0.972	0.779	1.556	1	0.212	0.378	[0.082, 1.743]
Middle (ref. high)	−0.614	0.690	0.794	1	0.373	0.541	[0.140, 2.090]
Educational level (overall test)	—	—	4.395	3	0.222	—	—
Primary/secondary education or below (ref. postgraduate)	−1.490	0.866	2.962	1	0.085	0.225	[0.041, 1.230]
High school (ref. postgraduate)	−0.917	0.522	3.082	1	0.079	0.400	[0.144, 1.113]
University (ref. postgraduate)	−0.632	0.460	1.888	1	0.169	0.532	[0.216, 1.309]
Employment status (overall test)	—	—	4.296	3	0.231	—	—
Employed (ref. student)	0.805	0.413	3.790	1	0.052	2.236	[0.995, 5.028]
Unemployed (ref. student)	0.811	0.481	2.845	1	0.092	2.250	[0.877, 5.775]
Retired (ref. student)	1.500	1.528	0.964	1	0.326	4.482	[0.224, 89.597]
CTQ emotional abuse	0.058	0.052	1.222	1	0.269	1.059	[0.956, 1.173]
CTQ physical abuse	0.071	0.053	1.774	1	0.183	1.074	[0.967, 1.192]
CTQ physical neglect	−0.064	0.053	1.479	1	0.224	0.938	[0.846, 1.040]
CTQ emotional neglect	−0.014	0.049	0.082	1	0.775	0.986	[0.895, 1.086]
CTQ sexual abuse	0.196	0.045	18.858	1	<0.001	1.217	[1.114, 1.330]
Extraversion	0.009	0.051	0.033	1	0.855	1.009	[0.913, 1.116]
Agreeableness	−0.084	0.048	3.036	1	0.081	0.919	[0.836, 1.011]
Conscientiousness	−0.055	0.052	1.088	1	0.297	0.947	[0.854, 1.049]
Openness to experience	−0.005	0.044	0.013	1	0.908	0.995	[0.913, 1.085]
Neuroticism	−0.044	0.026	2.780	1	0.095	0.957	[0.910, 1.008]
General self-efficacy	−0.034	0.017	3.909	1	0.048	0.966	[0.934, 1.000]
Relationship dependency	−0.002	0.016	0.014	1	0.907	0.998	[0.967, 1.030]

Note. Dependent variable: 0 = non-user, 1 = user. OR = odds ratio; CI = confidence interval; SE = standard error; and ref. = reference category.

3.8. Gender-Stratified Sensitivity Analyses and Multicollinearity Diagnostics

The gender-stratified analyses substantially qualified the unadjusted trauma findings. Among women (n = 229; 45 users and 184 non-users), the total childhood trauma score did not differ significantly between users and non-users (M = 43.13, SD = 14.37 vs. M = 39.51, SD = 14.40), Welch t(67.27) = 1.52, p = 0.134, d = 0.25. Childhood sexual abuse, however, remained higher among female users (M = 7.91, SD = 4.64) than female non-users (M = 6.18, SD = 3.13), Welch t(54.19) = 2.37, p = 0.022, d = 0.50. Among men (n = 202; 154 users and 48 non-users), both total childhood trauma and sexual abuse showed markedly larger differences: total trauma, M = 67.53 (SD = 15.69) versus M = 41.27 (SD = 11.35), Welch t(107.89) = 12.69, p < 0.001, d = 1.78; sexual abuse, M = 13.00 (SD = 4.71) versus M = 5.13 (SD = 0.64), Welch t(169.70) = 20.14, p < 0.001, d = 1.90.

Additional distributional and sensitivity analyses were conducted to examine the unusually large trauma-related differences among male participants. Among male non-users (n = 48), childhood sexual abuse scores showed a pronounced floor effect: 46 participants (95.8%) obtained the minimum possible score of 5, compared with 25 participants (16.2%) among male users (n = 154). The median sexual abuse score was 5 (interquartile range [IQR]: 5–5) among male non-users and 14 (IQR: 11–17) among male users.

To assess the robustness of the observed differences, supplementary Mann–Whitney U tests and Welch tests based on 20% trimmed means were conducted. Both approaches confirmed statistically significant differences between male users and non-users in total childhood trauma and childhood sexual abuse (all p < 0.001). These findings indicate that the observed group differences were not dependent solely on conventional parametric mean comparisons. Nevertheless, the pronounced floor effect among male non-users, the unequal group sizes, and the non-probability sampling strategy warrant caution when interpreting the magnitude and generalisability of these associations (Table 8).

Table 8. Gender-Stratified Comparison of Childhood Trauma by Online Dating Application Use Status.

Measure	Women Users, M (SD)	Women Non-Users, M (SD)	Women d (p)	Men Users, M (SD)	Men Non-Users, M (SD)	Men d (p)
Total childhood trauma	43.13 (14.37)	39.51 (14.40)	0.25 (0.134)	67.53 (15.69)	41.27 (11.35)	1.78 (<0.001)
Emotional abuse	9.62 (4.85)	8.14 (4.10)	0.35 (0.064)	13.48 (3.94)	8.10 (3.46)	1.40 (<0.001)
Physical abuse	5.89 (2.00)	6.10 (2.84)	−0.08 (0.568)	12.96 (4.57)	6.56 (3.09)	1.50 (<0.001)
Physical neglect	7.07 (2.55)	7.60 (3.45)	−0.16 (0.249)	13.66 (4.42)	9.10 (3.89)	1.06 (<0.001)
Emotional neglect	12.64 (5.00)	11.49 (4.99)	0.23 (0.169)	14.43 (3.44)	12.38 (4.48)	0.55 (0.005)
Sexual abuse	7.91 (4.64)	6.18 (3.13)	0.50 (0.022)	13.00 (4.71)	5.13 (0.64)	1.90 (<0.001)

Note. Welch-adjusted comparisons are reported separately by gender. d = Cohen’s d; values in parentheses in the d columns are p values.

Interaction analyses confirmed that the magnitude of the trauma difference varied by gender. After controlling for age, economic status, and educational level, the application-use-status × gender interaction was significant for total childhood trauma (B = 19.88, p < 0.001) and childhood sexual abuse (B = 5.73, p < 0.001). The adjusted use-status difference in total trauma was small and narrowly nonsignificant among women (B = 4.74, p = 0.051) but large among men (B = 24.63, p < 0.001). For sexual abuse, the adjusted difference remained significant in both women (B = 1.64, p = 0.010) and men (B = 7.38, p < 0.001), but was substantially larger among men.

Multicollinearity diagnostics for the final hierarchical logistic regression did not indicate a severe collinearity problem. Across the final predictor set, VIF values ranged from 1.26 to 4.59 and tolerance values ranged from 0.218 to 0.793. For the five childhood trauma subdimensions, VIFs ranged from 2.19 to 3.09; the sexual-abuse predictor had VIF = 2.19 and tolerance = 0.456. Thus, the independent sexual-abuse association in the final model cannot be attributed simply to extreme multicollinearity among the trauma predictors.

4. Discussion

In the present study, adults who used and did not use online dating applications were compared in terms of demographic characteristics, childhood trauma, Five-Factor personality traits, general self-efficacy, and relationship dependency, and the relative and incremental associations of these variables with application use were examined. The findings revealed three main patterns. First, application use was strongly associated with several demographic characteristics, particularly gender. Second, the most pronounced full-sample psychological differences emerged for childhood trauma; however, supplementary gender-sensitive analyses showed that the magnitude of these differences was not uniform across women and men. Third, when the variables were considered within the same hierarchical model, childhood trauma made the most substantial incremental contribution beyond demographic characteristics; among the trauma subdimensions, only childhood sexual abuse remained independently significant in the final model. Personality and general self-efficacy contributed relatively modest incremental information, whereas relationship dependency provided no additional explanatory value.

4.1. Demographic Characteristics

A strong association was identified between online dating application use and gender (φ = 0.57), with male participants showing a markedly higher rate of use than female participants. This finding is consistent with Castro et al. (2020), who reported a higher likelihood of application use among men. However, studies reporting no significant gender differences among users are also available in the literature (Holtzhausen et al., 2020). Systematic reviews likewise confirm this heterogeneity, indicating that being male may be associated with general application use or use for specific purposes, although gender differences may vary according to sample characteristics and motivations for use (Castro & Barrada, 2020; Konings et al., 2022). A recent study also reported that, among middle-aged users, men used a greater number of applications, had been users for longer periods, and used applications more frequently for casual sexual encounters than women (Jiménez-Muro et al., 2024).

The strong gender effect observed in the present study is particularly important for interpreting the user–non-user differences in psychological variables. The marked predominance of men in the user group and women in the non-user group raised the possibility that some unadjusted differences reflected group composition rather than a uniform association across gender. The supplementary sensitivity analyses confirmed this concern for total childhood trauma: the user–non-user difference was small and nonsignificant among women (d = 0.25, p = 0.134) but very large among men (d = 1.78, p < 0.001), and the adjusted application-use-status × gender interaction was significant. Accordingly, the large total-trauma effect in the full sample should not be interpreted as a gender-invariant population effect.

Age group, economic status, employment status, and relationship status were also significantly associated with application use. However, the effect sizes of these associations were smaller than that observed for gender. Although the online dating literature contains findings indicating that age, relationship status, and other sociodemographic variables are associated with the likelihood of use, results vary according to sample characteristics (Castro & Barrada, 2020; Konings et al., 2022). This supports the view that online dating application use should not be evaluated as a behavior independent of its sociodemographic context.

4.2. Childhood Trauma

One of the most pronounced findings of the study concerned childhood trauma. Participants who used online dating applications had markedly higher total childhood trauma scores than non-users, with a large effect size (d = 1.37). Users also had higher scores across all dimensions of emotional abuse, physical abuse, physical neglect, emotional neglect, and sexual abuse. The magnitude of the multivariate effect was also notable (η²p = 0.393).

These findings indicate that the association between online dating application use and adverse early-life experiences may not be reducible to a single type of maltreatment. However, this finding should not be interpreted as indicating that individuals who experience childhood trauma turn to online dating applications or that application use is a behavioral consequence of trauma. Most of the existing literature, rather than comparing users and non-users in terms of childhood trauma, has focused on the relationship of early maltreatment with motivations for use or online revictimization among individuals who already use these applications. Thus, the present findings extend the literature to a different level by indicating that an association may also exist between adverse early-life experiences and application use status.

At this point, it is important to consider the original adjusted main-effect analyses together with the supplementary gender-sensitive results. After controlling for age, gender, economic status, and educational level, the original multivariate model continued to show significant group differences in total childhood trauma and all trauma subdimensions. However, a main-effect adjustment for gender does not establish that the association is homogeneous across women and men. The supplementary interaction analyses demonstrated substantial gender heterogeneity for total childhood trauma and childhood sexual abuse. Thus, demographic adjustment reduced confounding but did not eliminate the need to examine effect modification by gender.

The pronounced floor effect in childhood sexual abuse scores among male non-users further complicates the interpretation of the exceptionally large effect size observed in men. Although supplementary nonparametric and trimmed-mean analyses supported the statistical robustness of the group difference, the restricted score distribution in this subgroup may have contributed to the magnitude of the observed standardized effect.

Thus, although childhood trauma remained associated with application-use status after adjustment for measured demographic characteristics in the overall sample, the supplementary gender-stratified and interaction analyses demonstrated that the magnitude of these associations differed substantially between women and men. More importantly, adding the childhood trauma subdimensions to the logistic regression after the demographic model increased Nagelkerke R² from 0.456 to 0.598 and produced a significant improvement in model fit (Δχ² = 76.242, p < 0.001). This result indicates that childhood trauma carries important information associated with application use status beyond demographic characteristics.

These findings are also consistent with the broader individual vulnerability perspective in the online dating literature. Recent systematic reviews and meta-analyses indicate small-to-moderate associations between application use and variables such as psychological distress, interpersonal sensitivity, and behavioral dysregulation, while the high heterogeneity across studies points to the importance of individual predispositions and social context (Cela & Wood, 2026; Sharabi et al., 2026). The present study extends this framework to early developmental experiences.

4.3. The Specific Role of Childhood Sexual Abuse

Among the trauma findings, childhood sexual abuse remained the most robust trauma correlate of application-use status. Sexual abuse showed the largest effect size among the trauma subdimensions in the full-sample comparison (η²p = 0.343), remained the most pronounced group difference after demographic adjustment (η²p = 0.155), and was the only trauma subdimension independently associated with application-use status in the final hierarchical logistic regression (OR = 1.217, 95% CI [1.114, 1.330], p < 0.001). The supplementary gender-stratified analyses strengthened but also qualified this finding: sexual-abuse scores were higher among users within both women (d = 0.50, p = 0.022) and men (d = 1.90, p < 0.001), while the significant interaction indicated that the magnitude of this association was substantially larger among men.

This finding substantially overlaps with the results reported by Pampanel et al. (2024). In that study, a history of childhood sexual abuse was associated with exposure to technology-facilitated sexual violence on dating applications, and users reporting childhood sexual abuse were also shown to differ in their motivations for use. Therefore, the importance of childhood sexual abuse in the context of online dating should be considered not only in relation to revictimization but also in terms of broader individual differences associated with digital romantic and sexual interactions.

Nevertheless, the specificity of the present finding should be interpreted cautiously. The loss of independent significance of the other trauma types in the final model does not mean that they are unrelated to application use, because significant full-sample differences were found across all trauma dimensions and the trauma measures were moderately interrelated. Importantly, the additional multicollinearity diagnostics did not indicate extreme collinearity: trauma-subdimension VIFs ranged from 2.19 to 3.09, and sexual abuse had VIF = 2.19 and tolerance = 0.456. Therefore, the sexual-abuse coefficient is not readily explained as an artifact of severe multicollinearity, although shared variance among trauma dimensions and residual confounding remain relevant.

Furthermore, because the study is cross-sectional, it cannot be claimed that childhood sexual abuse causes online dating application use, leads individuals to use these applications, or produces particular patterns of use. The findings indicate only an independent statistical association between the two variables after accounting for the other measured characteristics. Future longitudinal studies examining trauma history, attachment, expectations of intimacy, motivations for use, and in-app behaviors together are important for understanding the possible mechanisms underlying this association.

4.4. Personality Traits

In the unadjusted analyses, participants who used and did not use online dating applications differed significantly across all Five-Factor personality traits. Users had lower scores for extraversion, agreeableness, conscientiousness, neuroticism, and openness to experience. The most pronounced differences emerged for agreeableness (η²p = 0.213) and conscientiousness (η²p = 0.194). After controlling for demographic variables, the differences across all five personality dimensions remained significant; however, the effect sizes became considerably smaller, particularly for extraversion and neuroticism.

The most important aspect of these findings is the difference between the group comparisons and the multivariable logistic regression results. Adding personality traits as a block produced a small but significant incremental contribution to the model (Δχ² = 12.515, p = 0.028); however, none of the individual coefficients for the Five-Factor traits remained significant in the final model. Thus, although personality differed within the overall psychological profiles of users and non-users, the specific associations between individual personality traits and application use status weakened when demographic characteristics, childhood trauma, self-efficacy, and the other personality dimensions were considered simultaneously.

This pattern may help explain the inconsistent personality findings in the literature. Timmermans and De Caluwé (2017) reported user–non-user differences in some Five-Factor traits, whereas Castro et al. (2020), in a large university sample, found an association between openness and the likelihood of being a current user, with more limited findings for the other personality dimensions. Systematic reviews likewise indicate that the influence of personality may vary according to motivations for use rather than simply according to the binary distinction of “uses/does not use” an application (Castro & Barrada, 2020; Konings et al., 2022).

Accordingly, the present results indicate that caution is warranted when characterizing online dating application users in terms of a particular “personality profile.” Particularly when users and non-users differ substantially in demographic and developmental characteristics, concluding that users possess specific personality traits on the basis of unadjusted personality comparisons alone may be misleading.

4.5. General Self-Efficacy

Participants who used online dating applications had significantly lower levels of general self-efficacy than non-users, with a moderate effect size (d = 0.52). The difference remained significant after controlling for age, gender, economic status, and educational level. In addition, adding general self-efficacy to the hierarchical logistic regression made a small but significant contribution to the model (Δχ² = 4.198, p = 0.040), and self-efficacy showed a limited independent association in the final model (OR = 0.966, p = 0.048).

Although this finding indicates that general self-efficacy may be associated with application use, caution is warranted regarding the magnitude of the association. Adding self-efficacy increased Nagelkerke R² only from 0.619 to 0.626, and the upper bound of the 95% confidence interval for the final coefficient was 1.000. Therefore, this result, which lies at the threshold of statistical significance, should not be given the same weight as the associations observed for childhood trauma or gender.

Furthermore, the construct assessed in the present study was general self-efficacy rather than social self-efficacy, romantic relationship self-efficacy, or online communication self-efficacy. Therefore, the finding that users had lower general self-efficacy scores cannot be interpreted as indicating that they have inadequate face-to-face social skills or that they use online dating applications to compensate for social inadequacies. Although the existing literature has associated self-esteem, body image, and other self-evaluations with online dating behaviors, recent reviews indicate substantial conceptual and measurement heterogeneity in this field (Thomas et al., 2025). Thus, the present finding introduces general self-efficacy as an additional individual-difference variable into the online dating literature, while the specific mechanism underlying the association requires direct examination in future research.

4.6. Relationship Dependency

No significant difference in relationship dependency was found between participants who used and did not use online dating applications. The effect size in the unadjusted comparison was negligible (d = 0.07), and the result remained unchanged after controlling for demographic variables (η²p = 0.001). More importantly, adding relationship dependency at the final step of the hierarchical logistic regression provided no incremental explanatory value (Δχ² = 0.014, p = 0.907), and relationship dependency was not independently associated with application use in the final model (OR = 0.998, p = 0.907).

This finding indicates that online dating application use should not be directly equated with relationship dependency or dysfunctional interpersonal dependency. Dating applications are used for diverse motivations, including seeking romantic partners, socializing, sexuality, entertainment, curiosity, self-worth validation, and ease of communication (Konings et al., 2022). Therefore, merely using an application does not imply that an individual is excessively dependent on romantic relationships.

This result also supports the importance of the conceptual distinction between general use and problematic use. Recent systematic reviews indicate that the problematic online dating literature contains substantial heterogeneity in terms of definitions and measurement and that general use should not be regarded as an indicator of problematic use (Thomas et al., 2025). Nevertheless, because the Cronbach’s alpha coefficient of the Spann–Fischer Codependency Scale was 0.67 in the present sample, measurement error may have contributed to attenuation of the associations; therefore, the null finding should not be interpreted as definitive evidence of the absence of an association.

4.7. Incremental Explanatory Value and Integration of the Findings

The hierarchical logistic regression provides the principal finding of the study by evaluating the study variables within the same analytical framework. Whereas the first model containing only demographic characteristics had a Nagelkerke R² of 0.456, this value increased to 0.598 after the childhood trauma subdimensions were added. In contrast, adding the Five-Factor personality traits increased R² to 0.619, adding general self-efficacy increased it to 0.626, and adding relationship dependency did not alter the explanatory value. The final model achieved an overall classification accuracy of 82.8%, with a sensitivity of 75.4% and a specificity of 89.2%.

The stepwise pattern shows that the variable blocks differed in the amount of statistical information they added within this sample. Childhood trauma produced the largest increment after demographics, whereas personality and general self-efficacy produced smaller increments and relationship dependency produced none. This ordering should not be interpreted as a developmental sequence or explanatory mechanism. The blocks represent conceptually distinct domains entered in a prespecified order for incremental comparison; their changes in model fit do not demonstrate that trauma causes application use or mediates the effects of other variables.

It should nevertheless be emphasized that Nagelkerke R² does not have the same meaning as the proportion of explained variance in conventional linear regression. It is a pseudo-R² measure and should not be interpreted as indicating that the model “explains” exactly 62.6% of the variance in application use. The classification accuracy of 82.8% and ROC-AUC of 0.902 reflect the discrimination and classification performance of the original model within the study sample. Supplementary stratified five-fold cross-validation yielded a mean ROC-AUC of 0.873 and a mean classification accuracy of 79.12%, suggesting that discrimination remained relatively strong in held-out subsets of the available data. Nevertheless, because cross-validation employed a regularized estimator and all observations originated from the same non-probability sample, these findings should not be interpreted as independent external validation. Further validation in independently recruited samples is required to establish generalizability.

Taken together, the findings do not support conceptualizing online dating application use as a psychopathological behavior. Current meta-analytic evidence indicates that users and non-users differ on average in some domains of psychological health, but that the effects are generally small to moderate, substantial heterogeneity exists across studies, and causal mechanisms cannot be inferred from cross-sectional data (Cela & Wood, 2026; Sharabi et al., 2026). Accordingly, the present study treats online dating application use not as a risk behavior or an indicator of psychological disorder, but as a digital relationship behavior associated with various demographic, developmental, and psychological characteristics.

The principal contribution of the study is therefore narrower than a mechanistic account: it documents that retrospectively reported childhood adversity, broad dispositional traits, generalized self-efficacy, and relationship dependency do not contribute equally to distinguishing binary application-use status in this convenience sample. The particularly strong association involving childhood sexual abuse warrants independent replication, while its interpretation must remain sensitive to the methodological limitations discussed below.

4.7.1. Theoretical and Practical Implications

The findings support considering developmental history alongside demographic and individual-difference variables when studying digital romantic behavior, but they do not establish a developmental-relational pathway. Childhood trauma, personality, self-efficacy, and relationship dependency were included as complementary domains rather than as components of a validated causal model. The observed incremental association of childhood trauma indicates that trauma measures carried information about binary use status beyond the measured demographics in this sample; it does not identify why that association occurred.

Another theoretically important aspect of the findings is the distinction between unadjusted group differences and independent and incremental associations. Although users and non-users differed across all Five-Factor personality traits, none of these traits remained independently significant in the final logistic regression model. Similarly, general self-efficacy made a significant but limited contribution, whereas relationship dependency showed no incremental explanatory value. In contrast, adding childhood trauma produced a substantial improvement in model fit, and childhood sexual abuse remained independently significant among the trauma dimensions in the final model. This pattern suggests that multilevel models incorporating demographic and developmental factors may be more appropriate than reducing online dating application use to a single personality characteristic or a tendency toward interpersonal dependency.

From a practical perspective, it should be particularly emphasized that the findings do not provide grounds for characterizing online dating application users as a risky, dependent, or psychopathological group. The outcome variable assessed in the study was application use versus non-use, not problematic application use. Therefore, the observed associations cannot be interpreted as indicators of clinical diagnosis, dysfunction, or problematic use. Nevertheless, it may be useful for future studies to investigate the experiences of individuals with histories of childhood trauma in digital romantic and sexual interactions and to examine processes such as interpersonal trust, boundary setting, intimacy, rejection sensitivity, and technology-facilitated victimization.

4.7.2. Strengths and Limitations

This study has several strengths, including the simultaneous examination of multiple developmental and psychological domains, explicit reporting of effect sizes, demographic adjustment, and hierarchical modeling of incremental associations. These strengths, however, do not eliminate the consequences of the sampling and measurement design. In particular, statistical adjustment for gender and other measured demographics cannot make a non-probability sample representative or remove residual confounding from unmeasured variables.

Nevertheless, the findings should be evaluated in light of several limitations. First, the cross-sectional design does not allow the temporal ordering or causal relationships among the variables to be determined. Although childhood trauma clearly precedes application use in temporal terms, the present design does not allow the conclusion that trauma history causes subsequent application use. Similarly, it cannot be concluded that lower general self-efficacy increases application use or that application use affects self-efficacy.

Second, all data were obtained through self-report. Retrospective childhood-trauma reports may be affected by recall and reporting processes, and the use of a common self-report method across psychological variables may contribute to shared method variance. Third, convenience and snowball recruitment through Instagram, WhatsApp, LinkedIn, and Telegram creates a plausible digital-engagement selection mechanism: individuals who are more active in online environments may have been more likely to encounter or complete the survey and may also differ in their exposure to dating applications. Consequently, the sample should not be treated as representative of adults in Türkiye.

Fourth, the primary outcome in the present study was defined as a binary classification of online dating application use (0 = non-user; 1 = user). Although supplementary information regarding duration of use, daily time spent using applications, and motivations for use was collected from application users, these characteristics were not incorporated into the present analyses, which were specifically designed to compare users and non-users. Consequently, the binary outcome does not differentiate between occasional and intensive use, different motivations, specific application platforms, or distinct patterns of engagement. This limits the theoretical interpretation of the findings: the observed associations indicate differences in psychological characteristics between individuals classified as users and non-users, but they do not explain the purposes, intensity, or psychological processes underlying application use. The findings therefore cannot be generalized to specific motivations, problematic use, or particular forms of digital romantic behavior.

Fifth, the internal consistency coefficients for relationship dependency (α = 0.67) and openness to experience (α = 0.63) were relatively low. This increases measurement error and may attenuate associations, particularly null or weak findings. The absence of an independent relationship-dependency association and the weak role of openness should therefore not be interpreted as definitive evidence that these constructs are unrelated to application use.

Sixth, the user and non-user groups were markedly imbalanced by gender: 76.2% of men but only 19.7% of women reported application use (φ = 0.57). Although gender was included in adjusted models, covariate adjustment alone cannot resolve possible recruitment differences between the groups. The supplementary stratified and interaction analyses showed that this imbalance materially affected the magnitude of the trauma findings: the total-trauma difference was concentrated largely among men, whereas the sexual-abuse difference remained present in both genders but was substantially stronger among men. These results improve transparency but do not substitute for replication in a more gender-balanced probability-based or otherwise better-controlled sample.

Seventh, several potentially relevant confounders or explanatory variables, including sexual orientation, attachment style, depressive and anxiety symptoms, loneliness, self-esteem, religiosity, urban versus rural residence, detailed relationship history, psychiatric symptoms, and substance use, were not assessed in the present study. In addition, although application-use motivations were collected for a separate research question, they were not included in the present analyses. Consequently, childhood trauma cannot be interpreted as an isolated explanation of application-use status, as the observed associations may partly reflect broader psychosocial vulnerability, unmeasured determinants, or differences in motivations for use. Residual confounding therefore cannot be excluded, and the magnitude or direction of these potential influences cannot be quantified from the analyses reported here.

Finally, the unusually large trauma effect sizes in the full sample (d = 1.37 for total childhood trauma and η²p = 0.343 for sexual abuse in the unadjusted analyses) should not be interpreted as population estimates. The gender-stratified findings demonstrate that these pooled effects mask substantial heterogeneity, particularly the very large differences observed among men. Their magnitude may reflect genuine subgroup differences, but selection bias, recruitment through digitally engaged networks, group composition, retrospective self-report, and unmeasured confounding may also contribute.

5. Conclusions

In this cross-sectional convenience sample, adults classified as online dating application users and non-users differed in several demographic, developmental, and psychological measures. The most pronounced observed pattern involved retrospectively reported childhood trauma. However, the binary outcome, pronounced gender imbalance, non-probability social-media recruitment, self-report measurement, and unmeasured confounding substantially constrain the interpretation and generalizability of these differences.

Within the hierarchical logistic regression, adding childhood trauma after measured demographic variables improved model fit more than the later psychological blocks, and childhood sexual abuse remained independently associated with use status in the final model. Multicollinearity diagnostics did not indicate severe instability of this coefficient. However, supplementary gender-sensitive analyses showed that the broader total-trauma difference was largely concentrated among men and that even the sexual-abuse association was substantially stronger in men than in women. The central finding should therefore be understood as a sample-specific association with meaningful gender heterogeneity, not as evidence of a uniform developmental pathway to online dating application use.

The results therefore support a cautious, non-pathologizing conclusion: binary dating-app use status was associated with a combination of measured demographic and developmental characteristics, while personality, self-efficacy, and relationship dependency showed more limited independent contributions. The study does not establish a developmental mechanism and cannot determine whether the trauma associations reflect direct relationships, residual confounding, broader psychosocial vulnerability, or selection processes. The specific contribution of the present study is to clarify how childhood trauma and selected psychological characteristics are associated with the broad distinction between individuals who report using online dating applications and those who do not, rather than to explain the motivations or behavioral processes underlying application use. Accordingly, these findings should be interpreted as evidence of sample-specific psychological differences associated with use status, not as an explanation of why, how frequently, or for what purposes individuals engage with dating applications.

Future studies should use larger and more demographically balanced samples; record sexual orientation, relationship history, psychiatric symptoms, substance use, attachment, loneliness, self-esteem, and other relevant confounders; and distinguish application platform, current/past use, frequency, duration, intensity, and motivation. Gender-stratified analyses and independent replication of the childhood-trauma findings are particularly important. Longitudinal designs will be needed before developmental or causal pathways can be evaluated.